Erythropoietin receptor transcription is neither elevated nor predictive of surface expression in human tumour cells

Erythropoietin receptor (EpoR) has been reported to be overexpressed in tumours and has raised safety concerns regarding the use of erythropoiesis-stimulating agents (ESAs) to treat anaemia in cancer patients. To investigate the potential for EpoR to be overexpressed in tumours, we have evaluated human tumours for amplification of the EPOR locus, levels of EPOR transcripts, and expression of surface EpoR protein. Gene amplification analysis of 1083 solid tumours found that amplification of the EPOR locus was rare with frequencies similar to other non-oncogenes. EPOR transcript levels in tumours and tumour cell lines were low in comparison with bone marrow and were equivalent to, or lower than, levels in normal tissues of tumour origin. Although EpoR mRNA was detected in some tumour lines, no EpoR could be detected on the cell surface using 125I-Epo binding studies. This may be due to the lack of EpoR protein expression or lack of cell-surface-trafficking factors, such as Jak2. Taken together, we have found no evidence that EpoR is overexpressed in tumours or gets to the surface of tumour cells. This suggests that there is no selective advantage for tumours to overexpress EpoR and questions the functional relevance of EpoR gene transcription in tumours.

Erythropoiesis-stimulating agents (ESAs), epoetin-a, epoetin-b, and darbepoetin-a, regulate erythropoiesis by binding and activating the erythropoietin receptor (EpoR) on the surface of erythropoietic precursor cells, stimulating their survival, proliferation, and differentiation into mature red blood cells. Recently, offlabel investigational clinical trials that aimed to prevent anaemia in breast and head and neck cancer patients undergoing chemotherapy and radiotherapy, respectively, reported a decrease in locoregional progression-free survival and overall survival in ESA-treated patients (Henke et al, 2003;Leyland-Jones et al, 2005). These findings have led to speculation that administration of ESAs to cancer patients may promote tumour growth through stimulating EpoR expressed on tumours. This hypothesis appeared to be supported by the literature, as some investigators have reported that tumour samples and cell lines transcribe the EPOR gene at high levels, that 90 -100% of primary human tumours overexpress EpoR protein, and that recombinant human Epo (rHuEpo) induced proliferative, survival and migration effects on tumour cell lines (reviewed by Osterborg et al, 2007;Sinclair et al, 2007). In contrast, most clinical studies have found no increase in tumour progression or decrease in survival when ESAs are administered to cancer patients (Bohlius et al, 2006). Furthermore, many studies have found tumour cell lines are unresponsive to ESAs, including proliferation and survival of tumour cell lines in vitro (Berdel et al, 1991;Mundt et al, 1992;Rosti et al, 1993;Selzer et al, 2000;Westphal et al, 2002), and, of most physiological relevance, tumour growth in vivo (Silver and Piver, 1999;Mittelman et al, 2001;Stuben et al, 2001Stuben et al, , 2003Thews et al, 2001;Golab et al, 2002;Pinel et al, 2004;Sigounas et al, 2004;Ning et al, 2005). Thus, there is conflicting evidence for the role of EpoR in tumour growth.
Concerns have been raised about the lack of specificity of antibodies used by many studies to detect EpoR. Most studies used anti-EpoR peptide polyclonal antibodies (C-20; Santa Cruz Biotechnology, Santa Cruz, CA, USA) in immunoblotting and/or immunohistochemistry (IHC) experiments (Elliott et al, 2006). We and others have found that polyclonal antibodies C-20 detected a 66-kDa protein that was substantially larger than EpoR (59 kDa) and was identified to be heat-shock protein HSP70 (Elliott et al, 2006;Brown et al, 2007;Ragione et al, 2007). Owing to crossreactivity with non-EpoR proteins (e.g., HSP70), C-20 was found unsuitable for IHC (Elliott et al, 2006;Brown et al, 2007). In addition, other anti-EpoR antibodies have been found to be nonspecific and insensitive and therefore inappropriate for EpoR detection (Della Ragione et al, 2007;Kirkeby et al, 2007;Laugsch et al, 2008;Sturiale et al, 2007).
Since the findings and conclusions from studies that used C-20 and other anti-EpoR antibodies to detect EpoR are inconclusive, we performed an extensive investigation of the potential for EpoR to be overexpressed in tumour cells through antibody-independent approaches.

Genomic amplification
DNA for ROMA, comparative genomic hybridisation (CGH) and Q-PCR was isolated from snap-frozen tumour samples using a proteinase K method. Representational oligonucleotide microarray analysis was performed as previously described (Lucito et al, 2003;van der Horst et al, 2005), and CGH was performed as previously described (Pei et al, 2002). Data were normalised using a modified global loss algorithm (Smyth and Speed, 2003) (data not shown), and amplicon boundaries were determined using the GLAD algorithm (Hupe et al, 2004) and assembled into a database (data on file at Amgen Inc.).
Quantitative PCR analysis of the EPOR locus was performed on DNA from 68 primary breast tumours in multiplex reactions with one of two sets of EPOR-specific primers and probes. Primer/probe set A amplified an EPOR fragment within exon 3 ( Figure 1): forward primer, 5 0 -CTTCGTGCCCCTAGAGTTGC-3 0 ; reverse primer, 5 0 -TGATGTGGATGACACGGTGAT-3 0 ; and probe, 5 0 -TCACA GCAGCCTCCGGCGCT-3 0 . Primer/probe set C amplified an EPOR fragment within exon 8 that encodes the epitope for the M-20 anti-EpoR antibody ( Figure 1): forward primer, 5 0 -TGCCAGCTTTGAG TACACTATCCT-3 0 ; reverse primer, 5 0 -GCTCAGGGCACAGTGTC CAT-3 0 ; and probe, 5 0 -CCCAGCTCCCAGCTCTTGCGTC-3 0 . Probes were labelled with 6-carboxyfluorescein (FAM) at the 5 0 end and 6carboxytetramethylrhodamine (TAMRA) at the 3 0 end. A primer/ probe set specific for a relatively invariant region of chromosome 6p22.2 (data on file at Amgen Inc.) was included in each multiplex reaction, as a non-amplified control: forward primer, 5 0 -GGTCTCT ATTTGCACTTGGCTGAT-3 0 ; reverse primer, 5 0 -TTTTCATTGTT GACCAAGCTAGACA-3 0 ; and probe, 5 0 -TAGGGCATACTGCCTG CATATTTCCTGCT-3 0 labelled with VIC at the 5 0 end and TAMRA at the 3 0 end. PCR products were quantified from standard curves generated with each primer/probe set using normal human genomic DNA (Novagen, Madison, WI, USA). Each 10 ml Q-PCR mixture contained 5 ml 2 Â PCR master mix (Eurogentec, San Diego, CA, USA), 150 nM each primer and probe, and 15 -20 ng genomic DNA. DNA amplification was performed using the ABI 7900 (Applied Biosystems, Foster City, CA, USA): denaturation at 951C for 10 min, followed by 40 cycles of 15 s at 951C and 1 min at 601C. Reactions were performed in duplicate and repeated. Data were analysed using SDS software (Applied Biosystems). The relative copy number of EPOR using each primer/probe set is the ratio of the copy number at the EPOR locus to the copy number at the control locus; the mean relative copy number is presented.

Quantitative RT-PCR
For laser-dissected tumour and stroma RNA, sections were cut and placed directly into extraction buffer from Picopure RNA Isolation kit (sunnythanol (2 min), Accustain (30 s), dH2O (1 min), 75% EtOH, 95% EtOH, 100% Evale; Arcturus/Molecular Devices, Sunnyvale, CA, USA) or onto glass slides, then into a slide box on dry ice. Sections were stained as follows: treated with 75% EtOH ( Â 2), dipped in xylene, and then air-dried. Laser dissection was performed on a Pixcell IIe System and RNA was extracted using PicoPure Isolation kit (Arcturus/Molecular Devices), according to the manufacturer's instructions. RNA was quantified using the Beckman DU640 Spectrophotometer at the following wavelengths: 260, 280, and 320 nm.

Expression microarray analysis
Standard cRNA labelling and array processing were conducted according to the Affymetrix manual (Affymetrix, Santa Clara, CA, USA). First-strand synthesis used 5 mg total RNA, 10 pmol T7-(dT) 24 primer, and Superscript II (Invitrogen). Double-stranded cRNA was purified using the MinElute Reaction Cleanup kit (Qiagen). Biotinylated cRNA was synthesised using the Bioarray High Yield RNA Transcript Labeling kit (Enzo Diagnostics, Farmingdale, NY, USA) for 6 h at 371C and purified using the RNeasy Mini kit (Qiagen). The cRNA was hybridised to the Affymetrix Human Genome U133 Plus 2.0 array containing five unique EPOR probe sets: 209962_AT, 209963_S_AT, 215054_AT, 37986_AT, and 396_F_AT, each identifying different subregions within EPOR exon 8. The arrays were washed using the EukGE_WS2v4_450 protocol on a GeneChip Fluidic Station 450 and scanned using the Affymetrix GeneChip Scanner 3000. Raw signal intensities that met the recommended quality criteria were imported into the gene expression analysis system, Rosetta Resolver 5.0 (Rosetta Bio-software, Seattle, WA, USA; http:// www.rosettabio.com/tech/default.htm). System processing, including normalisation, consisted of algorithms using an errormodelling approach (Rajagopalan, 2003). A conventional two-way analysis of variance (ANOVA) was performed to determine if EPOR was differentially expressed in tumour vs normal tissues. Normalised intensity values were log transformed to achieve homoscedacity and used for the ANOVA calculations. An ANOVA calculation was performed for each tissue type with disease state and probe set as the two factors. If a difference was found, a post hoc comparison using the Dunn -Sidak test was performed. Statistical calculations were made using Matlab version 6.5.1 (Mathworks Inc., Natick, MA, USA).

Western blot analysis
Cell lysates and immunoblots were prepared as described previously (Elliott et al, 2006). Membranes were blocked with 5% milk in TBS containing 0.05% Tween-20, and incubated with 1 : 1000 dilution anti-Janus kinase 2 (Jak2) antibody (Cell Signaling Technologies, Danvers, MA, USA) for 2 h at room temperature. The secondary antibody was anti-rabbit IgG conjugated to horseradish peroxidase (Amersham Biosciences, Piscataway, NJ, USA). ECL-Plus (Amersham Biosciences) was used for detection. Blots were stripped and reprobed with 0.25 mg ml À1 anti-cyclophilin B (Abcam, Cambridge, MA, USA) as a loading control.

Epo-binding assays
Recombinant human Epo-binding studies were performed as previously described (Elliott et al, 1997) with some modifications. Cells were grown to confluency, dissociated with Versene (EDTA/ EGTA), centrifuged, washed, and resuspended in binding buffer (RPMI, 1% BSA, 0.1% sodium azide, 50 mM HEPES, 10 mg ml À1 cytochalasin B), and 1 Â 10 6 cells in 80 ml were incubated with 133 pM 125 I-rHuEpo (specific activity, 2500 Ci mmol À1 ), with or without a 1500-fold excess (200 nM) unlabelled rHuEpo, for 2.5 h at 371C. Reactions were quenched with 700 ml ice-cold PBS/0.5% BSA, and then centrifuged at 5000 g. Surface-bound 125 I-rHuEpo was determined using a Packard Cobra II Auto Gamma Counter (Perkin-Elmer, Boston, MA, USA) and specific binding calculated as the difference in 125 I-rHuEpo binding in the presence and absence of excess rHuEpo. For UT7/Epo Scatchard analysis, varying amounts of cold ligand (10 pg to 5 ng) in 40 ml binding buffer were added to 10 ml 125 I-rHuEpo and 50 ml 5 Â 10 5 UT7/Epo cells. Each reaction was performed in duplicate. Reaction tubes were incubated at 371C for 1.5 -2 h. Cell-bound and free 125 I-rHuEpo were separated through 100% dibutyl phthalate oil in a fixed angle horizontal rotor at 12 000 g for 2 min at room temperature. The tubes were frozen in dry ice, the oil phase containing cell-bound 125 I-rHuEpo was extracted, and the amount of 125 I-rHuEpo was determined by scintillation counting. Nonspecific binding was determined by adding a 300-fold excess of cold rHuEpo to the reaction mix. Scatchard analyses were performed using Microsoft Excel.

Genomic amplification of EPOR
Representational oligonucleotide microarray analysis and CGH data from 1083 tumour samples from 15 different tumour types were analysed for amplification of the EPOR locus. Amplification of EPOR in amplicons o10 Mb was identified in less than 0.7% of tumours (n ¼ 8) and was similar to the frequency of amplification of other established non-oncogenic loci such as glyceraldehyde-3phosphate dehydrogenase (GAPDH), b-actin (ACTB), and bglucuronidase (GUSB) (Figure 2A). None of the tumours with amplified EPOR contained EPOR in amplicons o1 Mb. Oncogenic loci cyclin D1 (CCND1), HER2, and EGFR were amplified in 3.5 -6.7% of tumours and frequently contained amplicons o1 Mb (Figure 2A). Copy numbers of oncogenic loci were increased 49fold in some tumours ( Figure 2B). In contrast, the copy number of EPOR was increased to a maximum of 2 -3-fold, concordant with other non-oncogenic loci ( Figure 2B).
To further analyse the frequency of EPOR amplification, Q-PCR was performed on genomic DNA isolated from 68 primary breast tumours. Copy numbers of EPOR were between 0.5 and 1.5 in all tumour samples, except one tumour with a copy gain and one with a copy deletion ( Figure 2C). Both EPOR primer sets yielded concordant results. Taken together, these results demonstrated that the EPOR locus was rarely amplified in primary human tumours.

EPOR transcripts in tumours and tumour cell lines
To investigate EPOR overexpression through mechanisms other than genomic amplification, levels of EPOR transcripts in tumours and cell lines were compared with levels in normal tissues of tumour origin. Analysis of 24 normal tissues showed that EPOR was transcribed at high levels in bone marrow, medium levels in adrenal, and low levels in all other tissues ( Figure 3A). Comparative analysis of EPOR transcription in tumour vs non-tumour tissues determined that transcript levels in kidney, stomach, colon, and lung tumours and cell lines were lower than those found in corresponding normal tissues ( Figure 3B). Levels of EPOR transcripts were also not elevated in brain tumours relative to normal brain ( Figure 3C), colon and lung adenocarcinomas relative to patient-matched normal tissue ( Figure 3D), and laserdissected tumour epithelia from head and neck tumours relative to stroma (samples collected from three different regions from each tumour A -C, Figure 3E).
Levels of EPOR transcripts were further examined in microarray analyses of 121 tumour and 170 normal tissues. For each of the five probe sets for EPOR on the HG-U133 Plus 2.0 array, levels of EPOR transcripts in tumours were not significantly elevated above levels found in the normal tissue counterpart (Figure 4; data not shown). Interestingly, a statistically significant (Po0.05) decrease in levels of EPOR transcripts in kidney tumours, lung squamous cell carcinoma, lymphomas, and prostate adenocarcinoma relative to normal tissues was identified (Figure 4). This result was confirmed with five different probe sets. These data demonstrate that EPOR transcription is not elevated in tumour samples or cell lines compared with EPOR transcript levels in the normal, non-tumour counterpart.

EPOR transcript levels and surface expression in tumour cell lines
Since EPOR transcripts were detected in tumour cells, albeit at low levels, we investigated the association between transcript levels and expression of EpoR at the cell surface. The particular cell lines used in this study were selected because of reports that they responded to Epo. The six cell lines included megakaryoblastic leukaemia line UT7/Epo cells (known to be Epo responsive; Laugsch et al, 2008) as a positive control; a renal carcinoma line 769P (with extremely low levels of EpoR mRNA) (Elliott et al, 2006) as a negative control; and four cell lines in which rHuEpo has been reported to induce an in vitro response: breast carcinoma line MCF-7 (Acs et al, 2001), cervical carcinoma line HeLa (Acs et al, 2003;Pajonk et al, 2004), renal carcinoma line CAKI-2 (Westenfelder and Baranowski, 2000), and neuroblastoma line SHSY-5Y (Um et al, 2007). Relatively high levels of EPOR transcripts were observed in UT7/Epo cells, very low levels in MCF-7, HeLa, SHSY-5Y, and CAKI-2 cells, and negligible levels in 769P cells ( Figure 5A). While UT7/Epo cells bound high levels, none of the solid tumour cell lines bound detectable levels above background of 125 I-rHuEpo, ( Figure 5B). Scatchard analysis demonstrated that UT7/Epo cells expressed B11 700 cell-surface receptors per cell with a binding affinity of 68 pM ( Figure 5C). In contrast, 32D cells expressing murine EpoR expressed relatively low levels of cell-surface receptors (0.58 ± 0.10% total c.p.m.; B760 receptors per cell; data not shown), demonstrating the assay was sensitive. These data indicate that synthesis of EpoR mRNA in solid tumour cell lines does not predict surface expression; EpoR protein may not be synthesised in these solid tumour cell lines and/or may be trapped in the cytoplasm and may not be trafficked to the surface of the cell. Janus kinase 2 is required for surface expression of EpoR in haematopoietic cells (Huang et al, 2001). Variable levels of Jak2 expression were observed in cell lines ( Figure 5D), and lack of Jak2 in some may account for, in part, lack of surface EpoR. However, other cell lines (e.g., CAKI-2) expressed Jak2 protein but still no surface expression was observed ( Figure 5B). These data suggest that if EpoR protein is synthesised in these cells, it does not get to the surface of tumour cell lines at detectable levels due to potential limiting surface-trafficking factors, such as Jak2. The EPOR-specific primer/probe sets A and C were used to amplify EPOR fragments from exons 3 and 8, respectively. Breast tumour no. 29 had a gain in EPOR copy number (1.6-fold) and no. 50 had a deletion of one EPOR locus (0.4-fold).

DISCUSSION
effects and promote tumour progression in anaemic cancer patients (Henke et al, 2003;Leyland-Jones et al, 2005;Bohlius et al, 2006). The notion that EpoR was overexpressed in tumours and that rHuEpo enhanced tumour progression has been confounded by the finding that antibodies used to detect EpoR are nonspecific and additionally bind non-EpoR proteins, including HSP70 (Elliott et al, 2006;Brown et al, 2007;Della Ragione et al, 2007;Kirkeby et al, 2007;Laugsch et al, 2008;Ragione et al, 2007;Sturiale et al, 2007). Therefore, a recent study suggesting high-level EpoR expression in head and neck tumours correlated with tumour progression and worse survival in patients administered ESAs (Henke et al, 2006) most likely identified the well-know association of HSP70 and worse prognosis (Schmitt et al, 2007). Furthermore, there are a large number of studies that find that ESAs do not stimulate tumour progression in preclinical models (reviewed by Osterborg et al, 2007;Sinclair et al, 2007). Since specific anti-EpoR antibodies have yet to be identified, we investigated if EpoR was overexpressed in tumours by performing a systematic analysis of EPOR genomic amplification and transcription in more than 15 different primary tumour types, and EpoR transcript and surface EpoR expression analysis in representative tumour cell lines in which rHuEpo has been reported to induce responses (Westenfelder and Baranowski, 2000;Acs et al, 2001Acs et al, , 2003Pajonk et al, 2004;Um et al, 2007).
A common phenomenon of tumour formation is the amplification of proto-oncogenes such as HER2 (Parkes et al, 1990), EGFR (Reissmann et al, 1999), CCND1 (Szepetowski et al, 1992;Reissmann et al, 1999), and c-MET (Rege-Cambrin et al, 1992), which provide a selective advantage for tumour cell growth and survival through overexpression. Genomic analysis of 1083 tumours showed that the EPOR locus was amplified in o0.7% tumours, only in large amplicons, and less than 2 -3 times the normal copy number. In contrast, EGFR, CCND1, and HER2 were amplified in 3.5 -6.7% of tumours, often 49 times the normal copy number and frequently amplified in small amplicons indicative of the selective amplification of these genes. These data demonstrate that EPOR amplification is a rare event in solid tumours and not a primary driver of tumour formation and progression.
Although the EPOR locus was rarely amplified in tumours, it was possible that EPOR was overexpressed through other mechanisms. Here, we report the first comprehensive, quantitative analysis of EPOR transcript levels in multiple tumour types compared to tissues of tumour origin. In this analysis, we found that levels of EPOR transcripts in tumour samples and cell lines from more than 15 different tumour types were equivalent to, or lower than, levels in normal tissues. In concordance with our study, it was recently reported that levels of EPOR transcripts in prostate (Feldman et al,   2006) and head and neck tumours (Winter et al, 2005) were similar to levels in normal tissues. High levels of EPOR transcripts were reported in kidney tumours compared with normal kidney (Lee et al, 2005) and in melanoma cell lines compared with normal melanocytes (Selzer et al, 2000). However, these latter studies used small numbers of samples, and levels of transcripts were not quantified. In the present study, we examined 11 kidney and 9 melanoma tumour samples and found no increase in levels of EPOR transcripts relative to normal tissues. Most solid tumour cell lines we examined expressed EpoR mRNA; however, the levels were substantially lower compared to that found in Epo-responsive cells (UT7/Epo) or tissues (bone marrow). In addition, we were unable to detect EpoR on the cell surface of the same cells. Owing to non-specificities and insensitivity of antibodies available for EpoR protein detection (Elliott et al, 2006;Brown et al, 2007;Della Ragione et al, 2007;Kirkeby et al, 2007;Laugsch et al, 2008;Sturiale et al, 2007), we have been unable to determine EpoR protein expression in these lines. However, the lack of detectable surface binding of 125 I-Epo to intact cells suggests that EpoR protein levels may be low if expressed at all. Translocation of EpoR to the cell surface is an inefficient process (less than 1% of total cellular EpoR molecules are produced on the cell surface); thus, EPOR transcription and EpoR protein synthesis does not always lead to cell-surface expression of functional EpoR in tumour cells (Migliaccio et al, 1991;Neumann et al, 1993;Sawyer and Hankins, 1993;Hilton et al, 1995;Hermine et al, 1996;Kurten et al, 1996;Supino-Rosin et al, 1999). Thus, an alternative explanation may be inefficient transport of EpoR protein to the cell surface. Janus kinase 2 is essential for surface expression of EpoR; in haematopoietic cells, it binds EpoR in the endoplasmic reticulum, induces correct protein folding, and localises EpoR to the cell surface (Huang et al, 2001). Some of the cell lines investigated had no or very low Jak2 expression and may, in part, account for the lack of detectable EpoR on the surface in some cells if protein is synthesised. Interestingly, murine 32D-Epo-independent cells expressed Jak2 1 10 100 1000 10000 Skin malignant melanoma (9) Prostate adenocarcinoma (5) Normal prostate (13) Pancreas adenocarcinoma (3) Ovary adenocarcinoma (3) Normal ovary (1) Lymphoma (8) Normal lymph node (6) Lung squamous cell carcinoma (8) Lung small cell carcinoma (5) Lung large cell carcinoma (5) Lung adenocarcinoma (9) Normal lung (47) Kidney carcinoma (11) Normal kidney (40) Colon adenocarcinoma (10) Normal colon (36) Breast lobular carcinoma (4) Breast ductal carcinoma (41) Normal breast (27) Fluorescence intensity   (DaSilva et al, 1994) and EpoR, but EpoR was not detected at the cell surface (Migliaccio et al, 1991). This suggests that accessory proteins other than Jak2 may also be required to modulate surface expression of EpoR or may alter the affinity of EpoR for Epo (Dong and Goldwasser, 1993;Masuda et al, 1993;Nagao et al, 1993;Hermine et al, 1994). Recently, it has been hypothesised that an alternative receptor for rHuEpo, a heterodimer of EpoR and the bcommon chain, mediates tissue-protective effects in non-haematopoietic tissues (Brines et al, 2004). However, these data were not reproduced by other investigators (Um et al, 2007). Taken together, these data suggest that Epo responses in these cells, if any, may be weak.
It is possible that very low levels of surface receptor on the tumour cell lines we examined are present and they mediate a response to rHuEpo, thereby explaining the responses to rHuEpo reportedly detected by some investigators (Westenfelder and Baranowski, 2000;Acs et al, 2001Acs et al, , 2003Um et al, 2007). Differentiated neuroblastoma cell line SHSY-5Y was reported to express extremely low (B17) EpoR homodimers on the cell surface (Um et al, 2007). This level of surface receptor was reportedly sufficient to induce low-level signalling and survival responses (Um et al, 2007). A proliferative effect of rHuEpo was reported in MCF-7 cells (Acs et al, 2001), but seven other studies reported no proliferative effect in MCF-7 cells (Berdel et al, 1991(Berdel et al, , 1992Mundt et al, 1992;Rosti et al, 1993;Gewirtz et al, 2006;Li et al, 2006;Laugsch et al, 2008). Also, rHuEpo reportedly increased the survival of parental HeLa cells in one study (Acs et al, 2003) but not in another study (Pajonk et al, 2004). Renal carcinoma cell line CAKI-2 was reported to bind 125 I-Epo and proliferate in vitro a maximum of 2.5-fold (Westenfelder and Baranowski, 2000). Since most of these studies had technical concerns (e.g., performed in the absence of serum, lack of critical controls (e.g., vehicle to control for contaminating carrier proteins), use of suprapharmacologic doses of rHuEpo, modest proliferative responses for established cell lines), the results from these studies are incon-clusive and inconsistent. In one such study, the direct effect of rHuEpo to induce signalling and protect a rat mammary carcinoma cell line from Taxol-induced apoptosis in vitro did not translate to an effect on Taxol-induced tumour inhibition in vivo, despite a reported impact on signalling pathways in the tumour (Hardee et al, 2006). The conclusion that ESAs promote tumour cell growth is not supported by in vivo studies in rodent tumour models: in 23 published studies, there was no tumourpromoting effect of ESAs either alone or in combination with radiotherapy or chemotherapy (Sinclair et al, 2007). In some studies, ESAs were reported to enhance the efficiency of tumourablative therapy (Silver and Piver, 1999;Mittelman et al, 2001;Thews et al, 2001;Golab et al, 2002;Pinel et al, 2004;Sigounas et al, 2004;Ning et al, 2005) and induce tumour regression (Mittelman et al, 2001). Therefore, the most physiologically relevant tumour models that directly examined effects of ESAs in vivo do not support a tumour-promoting effect of ESAs.
In summary, we found no evidence that EpoR mRNA was overexpressed in the solid tumour types analysed in this study (through genomic amplification or other mechanisms). Erythropoietin receptor transcript levels were low and EpoR protein was not trafficked to the cell surface at detectable levels. These data suggest that primary tumour samples and tumour cell lines did not overexpress EpoR. Taken together, these results question the hypothesis that there is functional relevance to EpoR mRNA transcription in tumours.